INDEX

Explanations

The neuron fires on the numeric‐and‐punctuation markers that denote structured, list-style or technical document formatting (e.g. digits, decimal numbers, hyphens, periods, etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 继续访问

1.26

ين

0.73

िया

0.72

↵↵

0.69

");*/

0.66

 into

0.66

ylamine

0.65

<strong>

0.63

 博文

0.63

kati

0.63

POSITIVE LOGITS

↵

3.66

<0x0D>

1.20

0.94

<0x0B>

0.86

</img>

0.81

”、

0.78

0.73

0.72

<h2>

0.70

Activations Density 1.241%